Itch.io Game Page Creation Doubles as AI Uploads Rise

Pages are rising faster than engagement

Itch.io’s upload spike is outpacing player demand. The indie storefront says new game pages have doubled in two months, while downloads and purchases haven’t kept pace. That gap raises a practical question for a store built around independent releases: how much of the increase comes from AI generated games, and can players still find the work they’re looking for?

A chart itch.io posted on Bluesky shows daily new game pages climbing from about 3,000 on August 3 to regularly more than 6,000 by October 30. That’s a sharp rise in a short stretch. It roughly matches the increase in demos at Steam’s latest Next Fest, where the count went from around 4,000 to more than 8,000. The comparison gives the itch.io figures some scale, though the two measures describe different things: one tracks new pages on a storefront, the other demos in a festival.

Itch.io hasn’t directly attributed the increase to AI. But when users asked what was behind it, the storefront replied “take a guess” with a robot emoji, and responded to a question about “AI dreck” with a weary face. The implication is clear, even if the exchange doesn’t establish how much of the growth is AI generated. Social posts can signal what a platform suspects; they can’t substitute for a breakdown of what’s actually being uploaded.

The platform’s own figures point to a supply and demand imbalance. Downloads and purchases haven’t doubled, despite the jump in pages. That suggests many new uploads are attracting little attention, adding to the work of finding games without necessarily bringing more players to the storefront. More releases can mean more chances for an unusual project to find its audience, but only if the store’s search and recommendation tools can surface it.

AI tools have widened the pipeline

The rise follows the July 24 launch of Claude Opus 5, which the source describes as particularly capable at reverse engineering code. Anthropic’s model has since advanced to version 5.5. These tools have affected game development in several ways, from speeding up the decompilation of older games to raising concerns among developers that public demos could be copied. Those are distinct uses, but they point to the same change: software that once demanded specialist knowledge may now be easier to experiment with.

That doesn’t mean every fast made game is AI generated, or that AI use alone explains the upload spike. Lower barriers can bring more people into game making, while also making it easier to produce projects that are unfinished, derivative, or hard to distinguish from one another. The storefront’s figures show the volume. They don’t explain the motives or methods behind each page.

Anthropic has updated Claude to limit its ability to copy games directly, but the broader effect on AI generated uploads remains uncertain. At the time of the report, 131,929 of the 1,769,316 games listed on itch.io carried an “AI Generated” tag—about 7.5% of the catalog. That’s a meaningful share, but it captures tagged games, not necessarily every upload made with AI assistance. Tags depend on creators applying them, so the number is best read as a visible minimum rather than a complete account of AI’s role.

Moderation alone may not solve discovery

Some users have called for itch.io to ban AI generated games. AI researcher Mike Cook argued that a ban wouldn’t necessarily change the upload trend: enforcement would require staff or a public reporting system, and detecting AI use can be harder than it appears. A rule would also need a workable definition of what counts as AI generated, especially when creators may use tools for one part of development and make other parts themselves.

Valve has also questioned whether banning AI coding tools on Steam would work, though its position reflects a more hands off approach to storefront management. The comparison illustrates the limits of a simple policy response. Even if a platform sets a boundary, it still has to identify relevant content and decide what evidence is enough to act on it. That takes time and attention that could otherwise go toward improving how games are presented to players.

Letting the surge fade on its own is another possibility. Some creators may be posting vibe coded games for online attention and eventually move on. Yet the engagement figures suggest that many uploads already struggle to find an audience. If you're working in this space, the distinction between a temporary wave of experimentation and a lasting change in publishing volume matters: the first may pass, while the second would leave discovery tools under continuing pressure.

Discovery is the longer term test

For itch.io, the larger issue is whether more pages make the store harder to navigate for players and developers alike. This is more significant than it looks. A crowded catalog can make it harder for players to find promising work, and harder for smaller creators to reach anyone beyond their existing audience. The platform benefits when creators can publish freely, but that freedom carries less value if new releases disappear into a growing stream.

There’s no evidence here that downloads or purchases are falling; the reported problem is that they haven’t risen alongside uploads. That distinction matters. A flat response to a sharp supply increase points to a mismatch, but it doesn’t by itself show that players are leaving or that the store is failing. The next useful signal would be whether engagement begins to track the larger catalog, and whether players can still locate games they want.

And yet, a larger catalog isn’t automatically a worse one. More people can publish, and some games that might once have remained private experiments can now be found by an audience. The test is whether itch.io can preserve that access while making discovery manageable (and this is the part most people overlook). More pages. Fewer guarantees of attention.

If uploads keep growing while engagement stays flat, discovery will become the test: can the platform help people find games they want before the volume pushes them elsewhere? The answer will depend less on the headline count of pages than on whether the store can make a larger supply useful to the people browsing it. That’s the question the current figures leave open.